All news

What is WebMCP and how do you prepare your Magento webshop for AI agents?

AI agents are going to search, compare, and order on behalf of your customers. WebMCP determines how they do that on your own site. What does that mean for your Magento webshop?

Two colleagues look at an online store together on a laptop
Two colleagues look at an online store together on a laptop


Short answer: Google does not set WebMCP requirements for webshops. WebMCP is a proposed web standard, currently in an origin trial in Chrome, with which your webshop itself offers tasks to AI agents in the browser, such as searching, filtering, or preparing an order. There is no standard extension for Magento yet; you add WebMCP in the frontend, for example in a Hyvä theme.

Updated on November 5, 2026. WebMCP is still under development and subject to change. The sources are listed at the bottom.

What is WebMCP?

WebMCP stands for Web Model Context Protocol. It is a proposed web standard from the W3C's Web Machine Learning Community Group. Google released an initial preview in Chrome in February 2026. Since Chrome 149, you can participate in the origin trial as a website.

The idea: an AI agent no longer has to guess what a button or field does. Your page itself reports which tasks an agent can perform and which input belongs to them. Chrome calls such a task a tool, for example 'checkout' or 'filter_results'. WebMCP offers two ways to provide tools:

  • Imperative API: you register tools with JavaScript via navigator.modelContext.

  • Declarative API: you annotate existing HTML forms so that the browser turns them into a tool.

The tools run visibly on your own page. For sensitive actions, such as a purchase, you can first ask the user for confirmation.

Does my Magento webshop have to comply with WebMCP?

No. WebMCP is a proposal in a testing phase, not a requirement from Google. Chrome describes it as a progressive enhancement: your webshop continues to work normally for visitors and agents without WebMCP. The advantage lies in reliability. An agent calling a tool makes fewer mistakes than an agent reading your page and trying buttons.

WebMCP, MCP, UCP, and ACP: what is the difference?

The acronyms can get mixed up, but they solve different problems.

Standard

Where it runs

What it's for

Status

WebMCP

In the browser, on your own page

Letting agents perform tasks on your site, with the user present

Origin trial in Chrome, since Chrome 149

MCP (Model Context Protocol)

On a server, separate from the browser

Connecting AI assistants to systems and data, such as your webshop's backend

Community modules for Magento 2; Adobe Commerce MCP is marked 'Coming soon' by Adobe

UCP (Universal Commerce Protocol)

With Google, in AI Mode and Gemini

Checking out without leaving the AI environment

Since January 2026, initially for US retailers

ACP (Agentic Commerce Protocol)

In ChatGPT

Recommending products and checking out in ChatGPT

Since September 2025; scaled back by OpenAI in 2026

In short: UCP and ACP are about checking out on the Google or OpenAI platform. MCP connects an AI assistant to your backend. WebMCP keeps the customer on your own site, in your own checkout, with your own prices.

Why is WebMCP interesting for B2B?

In B2B, the customer is often logged in, with custom prices, a personalized catalog, and specific ordering permissions. WebMCP fits well there: the agent works within the user's session, on the page the user sees. Consider tasks such as:

  • Entering a list of part numbers via quick order.

  • Searching and filtering products by technical specifications.

  • Submitting a request for quote.

  • Preparing a shopping cart that the buyer confirms themselves.

Quick ordering by item number and quote requests are already included in the Hyvä B2B Suite. With WebMCP, you make those features usable for an agent without the agent needing to mimic your page.

How do you prepare your Magento webshop for WebMCP?

  1. Lay the foundation. Agents first read your pages and data. Ensure schema.org markup, complete product data from your PIM, and content that is readable without JavaScript. See our 8 tips for AI and SEO.

  2. Choose three to five tasks. Which tasks do your customers perform most often? Start there, for example searching, filtering, quick ordering, and requesting a quote.

  3. Annotate existing forms. Search, quick order, and quote requests are forms. With the declarative API, you offer them as a tool.

  4. Build more complex tools with JavaScript. For example, filtering by multiple attributes or filling a shopping cart. In a Hyvä frontend, you integrate with existing Alpine.js components.

  5. Ask for confirmation on sensitive actions. Always let the user confirm orders and quotes themselves.

  6. Test in Chrome. Enable WebMCP via chrome://flags/#enable-webmcp-testing, inspect your tools with Chrome DevTools, and try them out with the Model Context Tool Inspector extension.

  7. Measure the effect. See if agents use your tools and if your visibility in AI responses changes.

What can WebMCP not do yet?

Chrome itself lists three limitations. WebMCP is primarily intended for agents in a browser with a human present, not for agents operating without a browser. Complex interfaces require extra JavaScript to maintain page state. And an agent only sees your tools once it visits your site.

Our advice: experiment on a test environment now and don't roll it out broadly just yet. The standard may still change. Those who get their product data and forms in order now will be able to adapt quickly once WebMCP becomes more widely available.

Frequently asked questions

Is there a Magento extension for WebMCP?

We are not aware of any standard extension for WebMCP yet. There are community modules that add an MCP server to Magento 2. Those run on the server, not in the browser.

Does WebMCP only work in Chrome?

Chrome is the first browser with an implementation, currently as an origin trial. The goal of the standard is for any browser with agent capabilities to support the API.

Do I need WebMCP to appear in ChatGPT or Google's AI mode?

No. Whether you appear in AI responses depends on your content, your product data, and mentions on other sites. WebMCP is about what an agent can do on your site.

Does WebMCP work with Adobe Commerce and Hyvä?

WebMCP runs in the frontend, so in principle it works with any Magento version. With Hyvä, you add tools to existing Alpine.js components. In addition, Adobe is working on its own Commerce MCP for Adobe Commerce; that is listed as 'Coming soon' by Adobe.

Is WebMCP secure?

Tools run on your own page, in the user's session. A Permissions Policy defines where tools can be registered; by default only on your own domain, not in iframes from other domains. Chrome also publishes guidelines for secure tools.

Want to know which tasks in your webshop are suitable for AI agents? We'd be happy to take a look together. Schedule a call.

Sources

  • Chrome for Developers: WebMCP (updated October 7, 2026)

  • WebMCP explainer on GitHub (W3C Web Machine Learning Community Group)

  • W3C AIKR Community Group: WebMCP Technical Notes

  • Adobe Experience League: Commerce MCP overview

  • Epartment: AI and SEO in e-commerce: 8 tips

Check out our archive of digital awesomeness

Our archive of digital awesomeness

More news.